Mendelian Randomization analysis of DeepSEA generated data

library(MendelianRandomization)

data_dir = "../../dat"
seq_predictions = read.csv(file.path(data_dir, "means_and_uncertainties.csv"))
seq_predictions = subset(seq_predictions, seq_num == 1)

str(seq_predictions)

'data.frame':   3001 obs. of  6 variables:
 $ seq_num    : int  1 1 1 1 1 1 1 1 1 1 ...
 $ mut        : Factor w/ 3000 levels "0A","0C","0G",..: 2680 1 2 3 342 343 344 674 675 676 ...
 $ X_pred_mean: num  -5.63e-05 -1.27e-06 -8.51e-07 -1.21e-06 5.42e-07 ...
 $ X_pred_var : num  -1.27e-07 -1.61e-07 -1.61e-07 -1.61e-07 -1.61e-07 ...
 $ Y_pred_mean: num  -9.79e-04 -1.01e-04 -5.30e-05 -3.94e-05 1.14e-04 ...
 $ Y_pred_var : num  -0.000496 -0.00049 -0.00049 -0.00049 -0.00049 ...

bx <- unlist(seq_predictions["X_pred_mean"])
bxse <- unlist(seq_predictions["X_pred_var"])
by <- unlist(seq_predictions["Y_pred_mean"])
byse <- unlist(seq_predictions["Y_pred_var"])

MRInputObject <- mr_input(bx = bx, 
                          bxse = bxse, 
                          by = by, 
                          byse = byse)

EggerObject <- mr_egger(MRInputObject, robust = TRUE, alpha = .05)
EggerObject

MR-Egger method
(variants uncorrelated, random-effect model)

Number of Variants =  3001 
Robust model used.
------------------------------------------------------------------

------------------------------------------------------------------
Residual Standard Error :  11.490 
Heterogeneity test statistic = 56644576.7190 on 2999 degrees of freedom, (p-value = 0.0000)
I^2_GX statistic: 100.0%

mr_plot(MRInputObject)

install.packages("devtools")

Installing package into ‘/home/stephenmalina/R/x86_64-pc-linux-gnu-library/3.6’
(as ‘lib’ is unspecified)
trying URL 'https://cloud.r-project.org/src/contrib/devtools_2.2.1.tar.gz'
Content type 'application/x-gzip' length 372273 bytes (363 KB)
==================================================
downloaded 363 KB

* installing *source* package ‘devtools’ ...
** package ‘devtools’ successfully unpacked and MD5 sums checked
** using staged installation
** R
** inst
** byte-compile and prepare package for lazy loading
** help
*** installing help indices
*** copying figures
** building package indices
** installing vignettes
** testing if installed package can be loaded from temporary location
** testing if installed package can be loaded from final location
** testing if installed package keeps a record of temporary installation path
* DONE (devtools)

The downloaded source packages are in
    ‘/tmp/RtmpCSK8Wq/downloaded_packages’

library(devtools)

Loading required package: usethis

install_github("WSpiller/RadialMR")

Skipping install of 'RadialMR' from a github remote, the SHA1 (18a7a6e9) has not changed since last install.
  Use `force = TRUE` to force installation

plot(by/bx, bx/byse, xlim=c(min((by-2*byse)/bx), max((bx+2*byse)/bx)), ylim=c(0, max(bx/byse)))
for (j in 1:length(bx)) {
lines(c((by[j]-1.96*byse[j])/bx[j], (by[j]+1.96*byse[j])/bx[j]),
c(bx[j]/byse[j], bx[j]/byse[j]))
}

abline(h=0, lwd=1); abline(v=0, lwd=1)

library(RadialMR)
r_input <- format_radial(bx, by, bxse, byse)

Missing SNP IDs; Generating placeholders

funnel_radial(egger_radial(r_input, alpha = .05))

Weights not specified: Adopting modified second-order weights

Radial MR-Egger

             Estimate Std. Error  t value     Pr(>|t|)
(Intercept) 17.428069  1.6839374 10.34959 1.088551e-24
Wj           7.145093  0.3373847 21.17788 6.972256e-93

Residual standard error: 91.317 on 2999 degrees of freedom

F-statistic: 448.5 on 1 and 2999 DF, p-value: <2e-16
Q-Statistic for heterogeneity: 25008282 on 2999 DF , p-value: 0

 Outliers detected